KR20180117460A - Method for detecting issue based on trend analysis device thereof - Google Patents

Abstract

Various embodiments of the present invention relate to a method for detecting an issue based on trend analysis and an apparatus thereof. According to an embodiment of the present invention, the method for detecting an issue based on trend analysis includes the steps of: generating time series data for two or more categories based on inputted data; detecting an abnormal trend based on component data included in the time series data; and determining a part of a plurality of keywords designated in two or more time points as the issue, based on at least two or more time points at which the abnormal trend is generated. Accordingly, the present invention can improve the detection performance of an important issue.

Description

[0001] METHOD FOR DETECTING ISSUE BASED ON TREND ANALYSIS DEVICE THEREOF [0002]

Various embodiments of the present invention relate to a method and apparatus for detecting an issue based on trend analysis, and more particularly to a method and an apparatus for analyzing a trend of input data and detecting a part of the trend as an issue .

A method for analyzing internal data of companies considering external environment is being studied by collecting Internet information. For example, trends of customers are identified by Internet information and analyzed with the company's performance (eg, unit-level sales) to determine the factors of sales change, or to base the planning of new products. These activities have a great impact on realizing the concept of real-time enterprise (RTE).

In the public sector as well as the corporate environment, it is possible to establish an immediate response and policy strategy by analyzing trends on the Internet, such as the safety of the public and the consumer environment. Overseas, these kinds of projects are under the name of "Horizon Scanning".

What is important at this time is to analyze the information collected? The problem is to detect the issue. How real-time issue detection can be performed can be an important factor in real-time management.

Patent Document 10-2014-0139466 discloses a data crawling engine unit for crawling past and real-time data on the web, which can be attributes for prediction; A database unit for storing and managing sales statistic data from past to present and the crawled attribute data; A data set generation unit for generating a data set including a training set and a test set for prediction, the data set including a data preprocessing step for data of the database unit; A data mining engine unit for calculating a sales forecast value corresponding to a forecast target date based on the data set through selection and predetermined prediction models and algorithms; A sales forecast information providing unit for providing the sales forecast value to the manager; An attribute data manual input unit for inputting data predicted by the manager to affect a sales forecast and using the data as attribute data for prediction to improve the accuracy of the prediction; And a sales forecast automation device based on manager input.

However, there are still technical limitations in detecting issues arising from input data, analyzing the relevance of issues in a timely manner, and detecting important issues.

10-2014-0139466 (Patent Document 1)

According to various embodiments of the present invention, it is possible to provide an apparatus and method for detecting an issue based on trend analysis, which detects an occurrence of an issue through trend analysis of inputted data and detects an important issue among the generated issues have.

According to various embodiments of the present invention, it is possible to provide a method and apparatus for detecting an issue based on trend analysis, which detects a key issue and its related keywords through trend analysis of time series data based on input data.

According to an embodiment of the present invention, there is provided a method of generating time series data, the method comprising: generating time series data for two or more categories based on input data; Detecting an abnormal trend based on the component data included in the time series data; And determining, as an issue, a part of a plurality of keywords designated at the two or more time points based on at least two time points at which the abnormal trend occurs.

According to various embodiments, the component data includes information on at least one of a period component, a trend component, an arbitrary component, and a moving average component of each of the time series data for each of the time series data .

The abnormal trend may be a state in which the value corresponding to the time component in the time series data of at least one of the two or more categories is out of a specified numerical range.

The step of detecting the abnormal trend may detect at least one of abnormality detected in the time series data of at least one of the two or more categories and abnormality detected.

The step of determining a part of the plurality of keywords designated at the two or more viewpoints as the issue may include applying at least one weight among the time weight according to the distance between the at least two viewpoints and the identity weight according to the identity of the category You can decide.

According to an embodiment of the present invention, there is provided a communication apparatus comprising: a communication unit for communicating with at least one external device; And generating a time series data for at least two categories based on data input from the external device through the communication unit, detecting an abnormal trend based on the component data included in the time series data, And a processor for determining a part of the plurality of keywords specified at the two or more viewpoints as an issue based on the two or more viewpoints.

According to various embodiments, the component data includes information on at least one of a period component, a trend component, an arbitrary component, and a moving average component of each of the time series data for each of the time series data .

In addition, the processing unit may determine, as the abnormal trend, a state in which a value corresponding to a time component in a time series data of at least one of the two or more categories is out of a specified numerical range.

In addition, the processing unit may detect at least one of the abnormality detected in the time series data of at least one of the two or more categories and the abnormal abnormality detected as the abnormality trend.

The processing unit may determine at least one of a plurality of keywords designated at the two or more time points as an issue by applying at least one weight among time weights according to distances between the at least two view points and identity weights according to the identities of the categories .

According to various embodiments of the present invention, the detection performance of important issues can be improved by analyzing time series data for various categories based on the collected information (data).

In addition, according to various embodiments of the present invention, it is possible to effectively determine the importance of an issue generated in real time or in the past by determining an important issue based on the degree of influence of the issue detection time on a plurality of issues to be generated .

1 is a flowchart illustrating an issue detection method according to an embodiment of the present invention.
2 is a flowchart illustrating a trend detection method in the issue detection method according to an embodiment of the present invention.
3 is a flowchart illustrating an abnormal point extraction method in an issue detection method according to an embodiment of the present invention.
FIG. 4 is a graph illustrating weights of past issue occurrence times and time differences in the issue detection method according to an embodiment of the present invention.
5 is a graph showing attribute data constituting time series data and time series data generated based on data input in the issue detection method according to an embodiment of the present invention.
FIG. 6 is a graph illustrating moving average data processed based on time series data and time series data in the issue detection method according to an embodiment of the present invention.
7 is a graph illustrating sampled data processed based on time series data in the issue detection method according to an embodiment of the present invention.
FIG. 8 is a diagram illustrating the components of an issue detection apparatus according to an embodiment of the present invention.

Hereinafter, various embodiments of the present invention will be described in detail with reference to the accompanying drawings. In order to clearly illustrate the present invention, parts not related to the description may be omitted. In addition, the same reference numerals can be used for the same or similar components throughout the specification.

In various embodiments of the present invention, expressions such as 'or', 'at least one', etc. may denote one of the words listed together, or may represent a combination of two or more.

The terms used in various embodiments of the present invention are intended to illustrate a specific embodiment and are not to be construed as limiting the invention, for example, the singular forms "a," "an, May include the meaning of.

Hereinafter, according to various embodiments of the present invention, a method for detecting a hot issue among various issues generated based on trend (or trend) analysis (hereinafter referred to as an issue detection method) Device. The issue detection method can be processed through at least one electronic device. According to an embodiment, an electronic device (hereinafter, an issue detection device) that processes an issue detection method generates time series data (or raw time series data) from input data and generates an issue based on trends detected through time series data You can decide.

1 is a flowchart illustrating an issue detection method according to an embodiment of the present invention.

Referring to FIG. 1, in operation 101, the issue detection apparatus confirms data input. Here, the data input to the issue detection apparatus may include information on two or more categories such as a keyword exposed through (or appearing on) online, and sales information of a company. As described above, the input data is not limited to the two categories of keyword and sales information of the company, but may include news related to industry, weather, sports, search word, holiday, anniversary, audience rating, Information about two or more categories can be inputted as data in various ranges such as a post,

The data input to the issue detection apparatus can be input in real time or can be input in accordance with a specified period. Here, if the issue detection apparatus has no data input, it can maintain the standby state of data input.

In operation 103, the issue detection apparatus generates time series data. The issue detection apparatus can classify input data according to a category and generate time series data. According to an embodiment, when the keyword and the sales information are input, the issue detection apparatus can generate the time series data by separating the keyword category and the sales category. In addition, when generating the time series data for the keyword, the issue detection apparatus can generate time series data for each keyword. Alternatively, the issue detection apparatus may group two or more keywords to generate time series data in units of groups. Here, the generated time series data may include a value associated with each category corresponding to the time component.

In operation 105, the issue detection apparatus detects a trend. The issue detection apparatus can detect the trend based on the time series data. At this time, the issue detection apparatus can detect the trend based on at least one component (or component data) included in the time series data, or can detect the trend based on the sampling data of the time series data.

The issue detecting apparatus detects an abnormal trend (or issue) among the detected trends, for example, a second or third point or an abnormal increase / decrease point. According to one embodiment, the issue detection apparatus can extract an anomaly point for at least one component among various components constituting time series data. In addition, the issue detection apparatus can extract an anomaly point for the sampled data generated based on the time series data. Here, the above-described plurality of issue detection models, for example, an operation of extracting an anomaly point or an operation of extracting an anomaly point can be performed for at least one of a plurality of categories including input data.

Hereinafter, with reference to FIG. 2 and FIG. 3, various embodiments in which the issue detection apparatus extracts an anomaly point or anomalous enhancement point based on time series data can be described.

2 is a flowchart for explaining a trend detection method according to an embodiment of the present invention. As described above, the trend detection includes an operation of extracting an anomaly point using component data of time series data for at least one category or an operation of extracting an anomaly increase / decrease direction using a sample value of time series data . Hereinafter, the various embodiments of FIG. 2 may be performed with some operations of the trend detection method 105 described in FIG. At this time, the operation 201 may be performed after operation 103 of FIG.

In operation 201, the issue detection apparatus determines whether to decompose the time series. For example, if the issue detection apparatus performs an operation of extracting an anomaly point of time series data and performs an operation of extracting an anomaly point of time series data, operation 207 An operation of calculating a moving average can be performed.

In operation 203, the issue detection apparatus performs time series decomposition. For example, the issue detection apparatus can detect at least one component based on the values constituting the generated time series data. Here, the components included in the time series data may be provided as constituent data including at least one of, for example, a period (or seasonal) component, a trend component, and an arbitrary component.

The period component data may include a period component and a time component. Similarly, the trend component data may include a trend component and a time component, and the arbitrary component data may include an arbitrary component and a time component. That is, as shown in FIG. 5, each of the components constituting the time-series data is composed of values for time components like the time-series data 500, and is displayed as a graph, a table, or a table according to the passage of time .

Here, the period data may be configured to include mutually repeated varying values with one unit period of the time range in which the expansion and contraction periods are specified. In addition, the trend data can be configured to include values for fluctuations indicating a tendency to rise or fall with a slope. In addition, the random data may be configured to include temporal issues, issues that have occurred suddenly in the past, and values related to the time related to the issue.

In operation 205, the issue detection device extracts an anomaly point for at least a portion of the component data. For example, the issue detection apparatus can confirm whether or not values corresponding to time components for component data are located within a specified range. For example, the issue detection apparatus can determine a point at which a value out of a specified numerical value range among values included in arbitrary component data is detected as an abnormal point of arbitrary component data.

Likewise, the issue detection apparatus can determine a value out of the specified numerical range out of the values included in the period component data or the trend component data, or a point at which a value greater than or equal to the specified value is detected as the abnormal point of the component data. At this time, two or more abnormal points may be included in each component data. Here, the viewpoint is not limited to the moment (or time) at which the component value deviates from the specified numerical value range, and the component value may include a state outside the numerical value range or a time range thereof.

The issue detection apparatus can terminate the embodiment of FIG. 2 or perform the operation 107 of FIG. 1 when the operation 205 is performed.

On the other hand, in operation 207, the issue detection apparatus calculates a moving average of time series data. The operation of calculating the moving average can be performed by a preprocessing operation for extracting an anomaly point using time series data.

The time series data generated based on the input values of the various categories can be configured to include abrupt variations. In order to improve the effect of processing nonlinear data, the issue detection apparatus can calculate a moving average at a designated time interval in time series data displayed in a graph.

In operation 209, the issue detection apparatus determines the increase / decrease width for the category based on the moving average result of the time series data. The issue detection apparatus can determine the increase / decrease width of the calculated moving average data so as to determine the anomalous increase / decrease point by using the moving average at the designated time intervals in the time series data.

According to one embodiment, the issue detection apparatus can perform sampling of time series data using moving average results of time series data. The issue detection apparatus can perform a derivative (e.g., a first derivative) on the moving average result of the time series data. At this time, the time series data may be nonlinear data, and therefore, the issue detection apparatus can numerically calculate the increase / decrease amount per unit time. For example, the sampled data may have a constant width (e.g., unit time) with respect to the time axis, and may include a value corresponding to each unit time.

In operation 211, the issue detection apparatus can extract an anomaly point of time-series data based on the calculated sampling data. For example, the issue detection apparatus can confirm whether values corresponding to time components separated by unit time of sampling data are located within a specified range. For example, if the issue detection apparatus detects a value out of a specified numerical range out of the values included in the sampled data, or a value greater than or equal to a specified numerical value, the issue detection apparatus can determine the point of time as an idealism point. In addition, when the issue detection apparatus confirms that the difference in the increase / decrease amount per unit time in continuous time is equal to or larger than the specified value, the issue detection apparatus can determine the point of time as an anomaly increase / decrease point.

According to one embodiment, the issue detection apparatus can determine the abnormal increase / decrease amount by comparing values corresponding to the unit time of the sampled data. For example, if the issue detection apparatus determines that the difference of the section distance values is larger than the designated numerical value (or a multiple, for example, 1.5 times) based on the section distance corresponding to the quarter quartile and the quarter quartile, The corresponding section can be determined as an abnormal increase / decrease point. At this time, two or more anomaly points detected based on the sampled data may be included. According to various embodiments, when the difference of the section distance value is larger than a specified value based on the section distance corresponding to 0 to 1/4 quartile and / or 3/4 to 1 quartile per unit time, Can be determined as an abnormal symptom.

Herein, it is obvious that the temporal position can be determined by various methods, as an example of a temporal position for describing an anomalous increase / decrease point determined based on a difference of values of a 1/4 quartile or a 3/4 quartile. In addition, according to the above description, the reference interval (or section distance) is described based on the quartile, but the present invention is not limited thereto, and various embodiments such as the 100th quartile, period, or numerical value are possible.

In addition, the issue detection apparatus is not limited to the method for determining the anomalous increase / decrease point described in operation 211, and it is possible to determine the anomalous increase / decrease point of the sampled data by applying at least some of various methods for determining the anomalous increase / decrease point Do.

The issue detection apparatus may terminate the embodiment of FIG. 2 or perform the operation 107 of FIG. 1 when the operation 211 is performed.

3 is a flowchart illustrating an abnormal point extraction method according to an embodiment of the present invention. Hereinafter, the various embodiments of FIG. 3 may be performed with some operations of the abnormal point extraction 205 described in FIG. At this time, the operation 301 may be performed after the operation 203 of FIG.

In operation 301, the issue detection apparatus can perform a regression analysis. The issue detection method according to an embodiment of the present invention may detect a feature of a sudden change, generation, or extinction of input data, and detect an issue based on the feature. At this time, the data-based features may be expressed in various components, but the arbitrary component data may include irregularly generated information among the input data. Therefore, the issue detection apparatus can perform a regression analysis on at least one component, for example, an arbitrary component included in the time-series data to determine a comparison group (or a comparison target group) for extracting an anomaly point.

The issue detecting apparatus can detect the trend component from the regression analysis result. Here, the component data of the raw time series data is defined as the raw component data, and the component data based on the regression analysis result can be defined as the last component data.

According to an embodiment for extracting an anomaly point by using a result of regression analysis, the last trend component data includes a trend for an arbitrary component. By comparing the raw arbitrary component data with the last trend component data, Can be detected

According to one embodiment, the issue detection apparatus can determine the difference between the raw arbitrary component data and the latest trend component data. Here, the " difference " may include the meaning of the residual excluding the same or similar values of the comparison object.

In operation 303, the issue detection device may determine an anomaly point based on a comparison result (e.g., a difference). According to one embodiment, the anomalous point may be determined based on a temporal reference specified for the difference between the last trend component data and the raw arbitrary component data. For example, the issue detection apparatus may detect a value larger than a specified multiple (for example, 1.5 times) of the value of the interval distance (d) based on the interval distance (d) between the quarter quartile and the quarter quartile, .

In addition, the issue detection apparatus is not limited to the method for determining the abnormal point described in the operation 303, but may be applied to various methods for determining the abnormal point, and at least some of the various methods for determining the abnormal increase / decrease point as described above It is obvious that an anomaly point based on the difference can be determined.

The issue detection apparatus may terminate the embodiment of FIG. 3 or perform the operation 107 of FIG. 1 when the operation 303 is performed.

Returning back to FIG. 1, the issue detection apparatus can determine a hot issue among the issues to be generated.

In operation 107, the issue detection apparatus determines an issue based on the detected anomaly point and the anomaly point based on the time series data. The issue may include at least a part of the keyword related to the point of time determined based on the anomaly point and the anomalous point of increase and the determined point in time. The issue detection apparatus may apply at least one issue judgment model in determining important issues based on the anomaly points and the anomaly points.

Here, the anomaly point and the anomalous increase / decrease point can be defined as the occurrence time of the issue. Further, in determining the issue, the issue detection apparatus can determine the issue by taking time elements into consideration. For example, when at least a part of the anomaly point and the anomalous increase / decrease point overlap within a specified time range from the point of occurrence of the issue, the issue detection apparatus confirms the duplication of the anomaly point and the anomalous increase / decrease point from the point of occurrence of the issue The probability of deciding the issue as an important issue can be handled by referring to the time distance.

According to an embodiment of the present invention, a comprehensive judgment model for determining important issues can be provided by the following equation (1).

(1)

(One)

는 검출된 이슈에 대한 종합판정값이다.

는 데이터의 카테고리로 입력되는 데이에 대하여 구분된 카테고리(

)에 대하여

=1, 2, 3…… 과 같이 설정될 수 있다.

는 현재 시점의 데이터 카테고리(

),

는 데이터의 카테고리에 따른 가중치(또는 가중치값)이다.

는 이슈가 감지된 현재 시점의 표본화 데이터 중 이슈감지 값이다.

는 과거

시점의 데이터 카테고리(

).

는 이슈가 감지된 과거

시점의 데이터 종류에 따른 가중치이다.

는 이슈가 감지된 과거

시점의 표본화 데이터 중 이슈감지 값이다. 또한,

는 현재시점과 과거

시점의 데이터 이질성에 대한 가중치로 수학식(2)로 제공될 수 있다.here,

Is a comprehensive determination value for the detected issue.

Is divided into categories (< RTI ID = 0.0 >

)about

= 1, 2, 3 ... ... As shown in FIG.

Is the data category of the current point (

),

Is a weight (or a weight value) according to a category of data.

Is the issue detection value among the sampling data at the present time when the issue is detected.

Past

Data category at point (

).

Is the past

It is the weight according to the data type of the viewpoint.

Is the past

It is the issue detection value among the sampling data of the viewpoint. Also,

The present time and past

(2) as a weight for the data heterogeneity of the viewpoint.

(2)

(2)

와 관련하여 과거 이슈 발생 시점과의 시간차에 대한 가중치 함수를 수학식(3)로 제공할 수 있다.Also,

(3), a weight function for the time difference with respect to the time when the past issue occurs.

(3)

(3)

는

이고,

는 현재 시점과 과거

시점의 시간차이고, 시간차는 일수로 제공될 수 있다.

는 강조지수로

시점에서의 최대 가중치이다.

는 최적 시간차로 중요 이슈로 결정하기 위한 최적의 시간차이다. 이때, 최적의 시간차는 기 설정된 상태이며, 특정 시점 또는 지정된 시간범위로 제공될 수 있다.

는 유효 이슈 시간범위이다.

는 현재시점의 이슈감지일로, 이슈를 감지한 현재 시점이며, 날짜 또는 시각을 포함하여 제공될 수 있다.

는 과거

시점의 이슈감지일이다.here,

The

ego,

The present time and past

Time difference, and the time difference can be provided in days.

As an emphasis index

It is the maximum weight at the time.

Is an optimal time difference for determining an important issue as the optimal time difference. At this time, the optimal time difference is preset and can be provided at a specific time point or a specified time range.

Is the effective issue time range.

Is a current issue detection date, a current time when an issue is detected, and may be provided including a date or a time.

Past

This is the issue detection date.

The issue detection apparatus can determine the comprehensive determination value c based on the equations (1) to (3), and when it is confirmed that the determined overall determination value c is equal to or larger than the specified value, .

The issue detecting apparatus can detect the keyword with respect to the determined important issue. For example, the issue detection apparatus can determine at least a part of the keywords having a high exposure frequency among the plurality of keywords included in the keyword category corresponding to the important issue as the keyword of the important issue. The issue detection apparatus can output the keyword of the determined important issue or important issue.

FIG. 4 is a graph 400 showing weights of past issue occurrence times and time differences in the issue detection method according to an embodiment of the present invention.

는 현재 시점에서

만큼의 과거시점(

에

의 가중치를 곱연산 처리하는 것으로 설명할 수 있다. 이때, 가중치의 크기는 도 4에 도시된 바와 같이 시간차가 커질수록 감소하고,

범위를 벗어나는 시점부터는 영향도가 사라지는 것으로 설명할 수 있다.Referring to FIG. 4 and the above-described equation (3)

At this point

Of past time points (

on

Can be explained by performing a multiplication operation process. At this time, the magnitude of the weight decreases as the time difference increases as shown in FIG. 4,

It can be explained that the influence disappears from the point outside the range.

)되고 과거

시점의 이슈(

)가 있다면, 수학식(3)에 의하여 가중치의 최대값(

)을 결정하여 종합판정값(c)이 높게 산출될 수 있다. 즉, 이슈의 재현이 특정 시간 간격을 갖는 경우 현재의 종합판정의 강도가 강화되는 것이다. 반면, 수학식(3)에 의한 가중치가 낮게 결정되더라도 이슈가 반복적으로 검출되는 경우 수학식(1)은 종합판정값(c)을 증가시키게 된다.That is, the issue detection apparatus can determine the current issue detection value considering the past issue occurrence time for a plurality of disparate categories by calculating the comprehensive judgment value (c) based on Equation (1). For example, if a specific issue is detected at the current point

) And past

Issue at the time (

), Then the maximum value of the weight (

), And the overall judgment value c can be calculated to be high. That is, when the issue reproduction has a specific time interval, the strength of the current comprehensive judgment is strengthened. On the other hand, when the issue is repeatedly detected even if the weight according to the equation (3) is determined to be low, the equation (1) increases the total judgment value c.

) 이후 유효 이슈 시간범위(

)에 도달하는 과정 상에서는 종합판정값(c)이 감소하며 결국 유효 이슈 시간범위(

)를 벗어나는 경우 해당 특정 이슈는 중요 이슈에 대한 영향력이 제거될 수 있다.In Equation (3), the time difference (

) Effective issue time range after (

), The total judgment value (c) decreases and eventually the effective issue time range

), The particular issue may have an impact on the critical issue.

Furthermore, as described above, the issue detection apparatus detects an issue arising for various categories by simultaneously processing disparate categories (e.g., d = 1, d = 2 ...) by determining important issues based on time , You can determine important issues related to multiple categories.

5 is a graph showing attribute data constituting time series data and time series data generated based on data input in the issue detection method according to an embodiment of the present invention.

The issue detection apparatus can selectively separate at least one component data included in the time series data 500. [ As shown in FIG. 5, the time series data 500 can be described by sorted data in which information of various categories inputted for the flow of time is sorted. The time series data 500 can be generated by receiving data input in real time or data for at least one time point for a specified time range.

Here, the horizontal axis of the time series data 500 may be provided as a time axis, and the vertical axis may be provided as a value for at least one category. Here, the time axis may include date and time information. Further, the vertical axis can collectively display information on two or more categories, or selectively display information of a specified category.

Based on the time series data 500, a plurality of component data can be detected. 5, the time-series data 500 may include periodic component data 510, trend component data 530, and optional component data 550. For example, The component data 510, 530, and 550 separated from the time series data 500 may be provided with a time axis for the horizontal axis and a value for at least one category for the vertical axis, as in the time series data.

FIG. 6 is a graph illustrating moving average data processed based on time series data and time series data in the issue detection method according to an embodiment of the present invention.

The issue detection apparatus can calculate the moving average of the time series data 610 and display the moving average data 630 corresponding to the time series data (for example, the graph 600). At this time, the issue detection apparatus can confirm the change or the flow of the numerical value for each category included in the time series data based on the moving average data. Here, at least one of a simple moving average, a weighted moving average, a geometric moving average, a cumulative moving average, and an exponent moving average may be applied to the moving average calculated through the issue detecting device.

Moving average data can also be provided as moving average values for at least one category, such as time series data, with the horizontal axis representing the time axis and the vertical axis representing the time axis data.

FIG. 7 is a graph illustrating sampled data processed based on time-series data in the issue detection method according to an embodiment of the present invention.

The issue detection apparatus may perform sampling based on moving average data 630. [ The sampling data may be formed to have a value in a unit time period as shown in FIG. The issue detection apparatus can confirm the amount of change of data based on the sampled data.

Sampling data can also be provided as a time axis for the horizontal axis and a value for at least one category as the time series data.

According to various embodiments of the present invention, the issue detection method may be performed through at least one device (e.g., issue detection device) as described above. Figure 8 illustrates the components of an issue detection device 800 in accordance with an embodiment of the present invention. The issue detection apparatus 800 includes at least one of a communication unit 801, an input unit 803, an output unit 805, a storage unit 807, and a processing unit 809.

The communication unit 801 can connect an external communication with the issue detection device 800. [ For example, the communication unit 801 is connected to a network through wireless communication or wired communication and can communicate with an external device. The communication unit 801 can receive data from at least one external device connected thereto. For example, the communication unit 801 can receive data related to at least one category for generating time series data from an external device.

The input unit 803 is a component for inputting information and / or a control command for processing data input to the issue detection apparatus 800 and includes, for example, a keyboard, a keypad, a touch screen, And a microphone. The input unit 803 generates input data in response to a user's operation.

The output unit 805 can output the processing result through the processing unit 809. [ For example, the output unit 805 may output at least a part of time series data, at least one component data constituting time series data, moving average data, sampling data, determined important issues, and information related thereto. According to one embodiment, the output 805 may include at least one of a display and a speaker.

When the output unit 805 is provided as a display, data transmitted to the display through the processing unit 809 may be displayed as a graphic user interface. When the output unit 805 is provided as a speaker, data transmitted to the speaker through the processing unit 809 can be output as audio.

The storage unit 807 can store instructions and / or data received or generated from the processing unit 809 or other components. The storage unit 807 may include, for example, a programming module such as a kernel, a middleware, an application programming interface (API), or an application. Each of the above-described programming modules may be stored in the storage unit 709 by software, firmware, hardware, or a combination of at least two of them.

In addition, the storage unit 807 stores data input through the communication unit 801 and / or the input unit 803, and stores data processed through the processing unit 809, keywords, time series data, at least one component data, At least some of the moving average data, sampling data, important issues and related information may be stored.

The processing unit 809 receives data from the above-described other components (for example, the communication unit 801, the input unit 803, the output unit 805, and the storage unit 807), confirms the received data, Processing of the confirmed data can be executed.

According to one embodiment, the processing unit 809 can confirm the data input through the communication unit 801 or the input unit 803. [ The processing unit 809 generates time series data for two or more categories based on input data, detects an abnormal trend based on the component data included in the generated time series data, and detects at least two or more abnormal trends A part of a plurality of keywords designated at the two or more time points may be determined as an important issue based on the time point.

Further, the processing unit 809 processes the data for at least one of the periodic component data, the trend component data, the arbitrary component data, and the moving average component data of each of the time series data for each of the time series data .

In addition, the processing unit 809 can determine, as an abnormal trend, a state in which a value corresponding to the time component in the time series data of at least one of the two or more categories is out of the specified numerical range.

Further, the processing unit 809 can determine at least one of the abnormality detected in the time-series data of at least one of the two or more categories and the abnormality detected as an abnormal trend.

The processing unit 809 applies a weight of at least one of a time weight according to the distance between at least two viewpoints and a uniformity weight according to the identity of the category to apply a part of the plurality of keywords specified at two or more time points to an important issue You can decide.

The processing unit 809 can also detect a keyword for the determined important issue. For example, the processing unit 809 can determine at least a part of the keywords having a high exposure frequency among a plurality of keywords included in the keyword category corresponding to the determined important issue, as the keyword of the important issue.

According to an embodiment of the present invention, the processing unit 809 may be divided into a plurality of logical or physical components. At this time, when the detailed components of the processing unit 809 are physically divided, each of the detailed components may be provided in a separate module form, or may be composed of one module in which two or more modules are connected.

The external apparatus connected to the issue detection apparatus 800 may be provided in the same or similar apparatus as the issue detection apparatus 800. [ In addition, the external apparatus may be configured as one or more of various apparatuses as in the issue detection apparatus 800. For example, the device may be provided as a structure that includes at least a portion of the devices described, or at least some of the functionality of the device.

As described above, the issue detection method according to various embodiments of the present invention can improve the detection performance of important issues by analyzing time series data for various categories based on the collected information (data).

In addition, the issue detection method according to various embodiments of the present invention determines important issues based on detection time of an issue for a plurality of issues to be generated, thereby not limiting the importance of real-time or past-generated issues, It is possible to effectively judge various information related to the generated issue or the determined important issue.

In addition, the issue detection method according to various embodiments of the present invention can effectively determine whether an important issue is important by determining the relevance of an issue generated at two or more viewpoints.

According to various embodiments, at least some of the devices and methods according to the various embodiments described in the claims of the present invention and / or the specification are in the form of hardware, software, firmware, or a combination of two or more of hardware, For example, modules, units).

A module may be a minimum unit or a portion thereof that performs various embodiments of the present invention as a minimum unit or a part of an integrally constructed component. The module may be implemented mechanically or electronically.

The embodiments of the present invention disclosed in the present specification and drawings are merely illustrative examples of the present invention and are not intended to limit the scope of the present invention in order to facilitate understanding of the present invention. Accordingly, the scope of the present invention should be construed as being included in the scope of the present invention, all changes or modifications derived from the technical idea of the present invention.

800: Issue detection apparatus 801:
803: input unit 805: output unit
807: Storage unit 809:

Claims (10)

Generating time series data for two or more categories based on input data;
Detecting an abnormal trend based on the component data included in the time series data; And
And determining a part of a plurality of keywords specified at the two or more time points as an issue based on at least two or more time points at which the abnormal trend occurred. The method according to claim 1,
Wherein the component data includes information on at least one of a periodic component, a trend component, an arbitrary component, and a moving average component of each of the time series data, for each of the time series data, based on a trend analysis Issue detection method. The method according to claim 1,
Wherein the abnormal trend is a state in which a value corresponding to a time component in time series data of at least one of the two or more categories is out of a specified numerical range. The method according to claim 1,
Wherein the step of detecting the abnormal trend detects at least one of an abnormal point and an abnormal symptom detected in time series data of at least one of the two or more categories. The method according to claim 1,
Wherein the step of determining a part of the plurality of keywords designated at the two or more viewpoints is determined by applying at least one weight among a time weight according to the distance between the at least two viewpoints and an identity weight according to the identity of the category , An issue detection method based on trend analysis. A communication unit for communicating with at least one external device; And
Wherein the control unit generates time series data for two or more categories based on data input from the external apparatus through the communication unit, detects an abnormal trend based on the component data included in the time series data, And a processor for determining a part of the plurality of keywords designated at the two or more viewpoints as an issue based on the above-mentioned viewpoints. The method according to claim 6,
Wherein the component data includes information on at least one of a periodic component, a trend component, an arbitrary component, and a moving average component of each of the time series data, for each of the time series data, based on a trend analysis Issue detection device. The method according to claim 6,
Wherein the processing section determines, as the abnormal trend, a state in which a value corresponding to a time component is out of a specified numerical range in time series data of at least one of the two or more categories. The method according to claim 6,
Wherein the processing section detects at least one of an abnormal point and an abnormal symptom loss detected in the time series data of at least one of the two or more categories as the abnormal trend. The method according to claim 6,
Wherein the processing unit determines at least one of a plurality of keywords specified at the two or more time points as an issue by applying at least one weight among a time weight according to the distance between the at least two view points and an identity weight according to the identity of the category, Issue detection device based on trend analysis.

Images